Virtual Twin Research Articles

Reinforcement learning and digital twin-based real-time scheduling method in intelligent manufacturing systems

Optimization efficiency and decision-making responsiveness are two conflicting objectives to be considered in intelligent manufacturing. Therefore, we proposed a reinforcement learning and digital twin-based real-time scheduling method, called twins learning, to satisfy multiple objectives simultaneously. First, the interaction of multiple resources is constructed in a virtual twin, including physics, behaviors, and rules to support the decision-making. Then, the real-time scheduling problems are modeled as Markov Decision Process and reinforcement learning algorithms are developed to learn better scheduling policies. The case study indicates the proposed method has excellent adaptability and learning capacity in intelligent manufacturing.

A Biorealistic Computational Model Unfolds Human-Like Compliant Properties for Control of Hand Prosthesis.

Objective: Human neuromuscular reflex control provides a biological model for a compliant hand prosthesis. Here we present a computational approach to understanding the emerging human-like compliance, force and position control, and stiffness adaptation in a prosthetic hand with a replica of human neuromuscular reflex. Methods: A virtual twin of prosthetic hand was constructed in the MuJoCo environment with a tendon-driven anthropomorphic hand structure. Biorealistic mathematic models of muscle, spindle, spiking-neurons and monosynaptic reflex were implemented in neuromorphic chips to drive the virtual hand for real-time control. Results: Simulation showed that the virtual hand acquired human-like ability to control fingertip position, force and stiffness for grasp, as well as the capacity to interact with soft objects by adaptively adjusting hand stiffness. Conclusion: The biorealistic neuromorphic reflex model restores human-like neuromuscular properties for hand prosthesis to interact with soft objects.

A Digital Twin concept for the prescriptive maintenance of protective coating systems on wind turbine structures

The application of protective coating systems is the major measure against the corrosion of steel for tower sections of wind turbines. The inspection, condition monitoring and maintenance of protective coating system is a demanding and time-consuming procedure and requires high human effort. The paper introduces for the first time a Digital Twin concept for the condition monitoring and prescriptive maintenance planning for surface protection systems on wind turbine towers. The initial point of the concept is an in-situ Virtual Twin for the generation of reference areas for condition monitoring. The paper describes the integration of an online image annotation and processing tool, a maintenance model, corrosive resistance parameters, structural load parameters, and sensor data into the Digital Twin concept. The prospects of the concept and its practical relevance are shown for tower structures of large onshore wind turbines.

Computational Approaches for Supporting Combination Therapy in the Post-Aducanumab Era in Alzheimer’s Disease

With the approval of aducanumab on the “Accelerated Approval Pathway” and the recognition of amyloid load as a surrogate marker, new successful therapeutic approaches will be driven by combination therapy as was the case in oncology after the launch of immune checkpoint inhibitors. However, the sheer number of therapeutic combinations substantially complicates the search for optimal combinations. Data-driven approaches based on large databases or electronic health records can identify optimal combinations and often using artificial intelligence or machine learning to crunch through many possible combinations but are limited to the pharmacology of existing marketed drugs and are highly dependent upon the quality of the training sets. Knowledge-driven in silico modeling approaches use multi-scale biophysically realistic models of neuroanatomy, physiology, and pathology and can be personalized with individual patient comedications, disease state, and genotypes to create ‘virtual twin patients’. Such models simulate effects on action potential dynamics of anatomically informed neuronal circuits driving functional clinical readouts. Informed by data-driven approaches this knowledge-driven modeling could systematically and quantitatively simulate all possible target combinations for a maximal synergistic effect on a clinically relevant functional outcome. This approach seamlessly integrates pharmacokinetic modeling of different therapeutic modalities. A crucial requirement to constrain the parameters is the access to preferably anonymized individual patient data from completed clinical trials with various selective compounds. We believe that the combination of data- and knowledge driven modeling could be a game changer to find a cure for this devastating disease that affects the most complex organ of the universe.

THE IMPACT OF VIRTUAL TOURS ON MUSEUM EXHIBITIONS AFTER THE ONSET OF Covid-19 restrictions: visitor engagement and long term PERSPECTIVES

After the outbreak of Covid-19, galleries and museums have been experimenting with new ways to engage a potential audience remotely. This study focuses on the level of engagement of virtual tours in museums looking at the representation of architectural space, representation artifacts, and ease of use as possible correlated factors. A sample group of eighty early-career experts in the field of art, architecture, or design assessed their visit to the archaeological museum of Troya Muzesi in Canakkale, Turkey;half of the participants resided in Turkey, while the other half in Italy. This paper has addressed the following research questions with an online multi-level study: how is the online exhibition platform evaluated by its audience? Can regular employment of virtual tours engage new visitors in the long term? Is the representation of a museum, in the form of a virtual twin, an adequate surrogate that creates an immersive visiting experience? © 2021. All Rights Reserved.

Subgroup discovery in non-inferiority trials.

Approaches and guidelines for performing subgroup analysis to assess heterogeneity of treatment effect in clinical trials have been the topic of numerous papers in the statistical and clinical literature, but have been discussed predominantly in the context of conventional superiority trials. Concerns about treatment heterogeneity are the same if not greater in non-inferiority (NI) trials, especially since overall similarity between two treatment arms in a successful NI trial could be due to the existence of qualitative interactions that are more likely when comparing two active therapies. Even in unsuccessful NI trials, subgroup analyses can yield important insights about the potential reasons for failure to demonstrate non-inferiority of the experimental therapy. Recent NI trials have performed a priori subgroup analyses using standard statistical tests for interaction, but there is increasing interest in more flexible machine learning approaches for post-hoc subgroup discovery. The performance and practical application of such methods in NI trials have not been systematically explored, however. We considered the Virtual Twin method for the NI setting, an algorithm for subgroup identification that combines random forest with classification and regression trees, and conducted extensive simulation studies to examine its performance under different NI trial conditions and to devise decision rules for selecting the final subgroups. We illustrate the utility of the method with data from a NI trial that was conducted to compare two acupuncture treatments for chronic musculoskeletal pain.

Global Optimal Experiment Design for Li-Ion Batteries

We propose a method that enables the design of parameter specific optimal experiments to complement existing parametrization approaches. Previous studies on optimal experiments focus on fisher information based experiment design [1][2], which employ local sensitivities to characterize experiment information content. This has the distinct weakness of dependency on the initial parameter vector. We propose a method utilizing a global sensitivity scheme to fully map the available parameter space and thus account for model non-linearity and parameter interdependence. Furthermore, common optimal experiment design approaches for physics-based battery models consider a limited experiment space, choosing from a set a pre-defined experiments [2]. We circumvent this issue by algorithmically generating virtual experiments and finding the optimal one within this space. We validate our approach using both model and experiments. The experimental validation is based on the comparison with conventional battery characterization tools such as electrochemical impedance spectroscopy (EIS), galvanostatic intermittent titration technique (GITT) or scanning electron microscopy (SEM).With the accelerating growth in battery electric vehicles (BEV), efficient utilization of rapidly evolving system designs and chemistries shifts into focus. Recent developments towards nickel-rich cathode materials aiming for higher energy densities can only be an intermediate step towards the next generation of batteries. These fast paced improvements pose an immense challenge for the conventional approach to battery management systems (BMS), mainly in their reliance on empirical and experimentally validated control oriented models. Trends are pointing towards physics based (PM) as well as data-driven models (DM) for BMS, whose accuracy and resolution enables aging sensitive control. PM rely on long term cycling studies and detailed electrochemical characterization procedures for accurate parametrization.We propose a method for model parametrization utilizing global sensitivity based optimal experiment design. The method improves on conventional techniques in four ways: firstly, it significantly reduces computational and experimental effort, as we show by studying the impact of test length on identifiability. Secondly, it improves model accuracy through reductions in parameter error. Thirdly, it improves model fidelity by specifically designing experiments to excite desired parameters and finally, it enables full-cell parametrization with small experimental effort. We propose several use-cases for this method: beginning of life (BOL) parametrization, aging-specific experiments and model reparametrization and finally virtual twin [3] parametrization using a DM surrogate. We investigate which qualitatively dependent parameters can be untangled and therefore accurately determined from full cell measurements. We benchmark our method with half-cell experiments of nickel-rich electrodes harvested at beginning of life from commercial cells.[1] A. Pozzi, G. Ciaramella, S. Volkwein, and D. M. Raimondo, “Optimal design of experiments for a lithium-ion cell: Parameters identification of an isothermal single particle model with electrolyte dynamics,” en, Industrial & Engineering Chemistry Research, vol. 58, no. 3, pp. 1286–1299, 2019[2] S. Park, D. Kato, Z. Gima, R. Klein, and S. Moura, “Optimal Experimental Design for Parameterization of an Electrochemical Lithium-Ion Battery Model,”, Journal of The Electrochemical Society, vol. 165, no. 7, A1309–A1323, 2018[3] B. Wu, W. D. Widanage, S. Yang, and X. Liu, “Battery digital twins: Perspectives on the fusion of models, data and artificial intelligence for smart battery management systems,”, Energy and AI, vol. 1, 2020 Figure 1

Towards virtual twin for electronic packages in automotive applications

The piezoresistive silicon based stress sensor has the potential to be part of the Digital Twin implementation in automotive electronics. One solution to enforce reliability in digital twins is the use of Machine Learning (ML). One or more physical parameters are being monitored, while other parameters are projected with surrogate models, just like virtual sensors. Piezo-resistive stress sensors are employed to measure the internal stresses of electronic packages, an Acquisition Unit (AU) to read out sensor data and a Raspberry Pi to perform evaluation. Accelerated tests in air thermal chamber are performed to get time series data of the stress sensor signals, with which we can know better about how delamination develops inside the package. In this study stress measurements are performed in several electronic packages during the delamination. The delamination is detected by the stress sensor due to the continuous change of the stiffness and the local boundary conditions causing the stresses to change. Although, the stress change in multiple cells can give enough information if it is delaminated or not, its delamination area location is unknown. Surrogate models built upon Neural Networks (NN) and Finite Element Method (FEM) are developed to predict the out of plane stresses at the delaminated layer. FEM simulation models are calibrated with Moiré measurements and validated at the component and PCB level with stress difference measurements. Simulation delamination areas are constructed based on the Scanning Acoustic Microscope (SAM) images, and are also validated with the equivalent stress measurements. In the end the surrogate model is predicting the out of plane stress in the adhesive layer. The results show good correlation when compared to the SAM images.

Cutting Parameter Optimization for Reducing Carbon Emissions Using Digital Twin

With the exacerbation of global environmental concerns, manufacturing industries need to consider the impact of carbon emissions from manufacturing processes. The selection of the parameters in the machining process greatly influences on carbon emissions and machining efficiency. Hence dynamically optimizing the machining process parameters is a significant means to reduce carbon emissions according to the real-time perception of the machining conditions. In the paper, a method of cutting parameter optimization is presented on basis of the construction the digital twin of a CNC machine tool. In this method, an ontology on CNC machining process is established to be used as a communication bridge for understanding the semantic of the real-time interaction between the physical machine and the virtual twin. And a dynamic optimization method on cutting parameters is presented according to the simulation and optimization of the virtual twin with the dynamic perception of the machining conditions of the physical machine. At last, a case study is presented to validate this method for effectively optimizing the cutting parameters and decreasing carbon emissions.

Virtual Electrode Design for Lithium‐Ion Battery Cathodes

Microstructural characteristics of lithium‐ion battery cathodes determine their performance. Thus, modern simulation tools are increasingly important for the custom design of multiphase cathodes. This work presents a new method for generating virtual, yet realistic cathode microstructures. A precondition is a 3D template of a commercial cathode, reconstructed via focused ion beam/scanning electron microscopy (FIB/SEM) tomography and appropriate algorithms. The characteristically shaped micrometer‐sized active material (AM) particles and agglomerates of nano‐sized carbon‐binder (CB) particles are individually extracted from the voxel‐based templates. Thereby, a library of roughly 1100 AM particles and 20 CB agglomerates is created. Next, a virtual cathode microstructure is predefined, and representative sets of AM particles and CB agglomerates are built. The following re‐assembly of AM particles within a predefined volume box works using dropping and rolling algorithms. Thereby, one can generate cathodes with specified characteristics, such as the volume fraction of AM, CB and pore space, particle‐size distributions, and gradients thereof. Naturally, such a virtual twin is a promising starting point for physics‐based electrochemical performance models. The workflow from the commercial cathode microstructure through to a full virtual twin will be explained and assessed for a blend cathode made of the two AMs, LiNiCoAlO2 (NCA) and LiCoO2 (LCO).

Application and Analysis of Random Forest Algorithm for Estimating Lawn Grass Lengths in Robotic Lawn Mower

This paper states an estimation method for lawn grass lengths or ground conditions based on random forest algorithm from the observation data obtained by fusion of sensors. This estimation relates to Digital Twin and Virtual Twin of Hybrid Twin approach for the autonomous driving of robotic lawn mowers. The robotic lawn mowers are becoming popular with the advent of efficient sensors and embedded systems and we are now developing a practical autonomous driving and its group control algorithm for large lawn grass areas. However, one of the important functions of robotic lawn mower, that is, the length of lawn grasses or such ground conditions as dirt, gravel, or concrete, etc., are not recognized precisely with the current robotic lawn mower. As a result, the motor for cutting lawn grasses is running with constant rotation speed from the beginning to the end of operation of robotic lawn mower. This leads to the waste of battery and gives a large drawback for the control of robotic lawn mower. In order to precisely control the rotation speed of motor and save the battery, the lawn grass lengths and ground conditions are estimated by using the effective sensor data. The application of random forest algorithm to the fusion of sensors on a commercial robotic lawn mower attained more than 90% correct estimation ratio in several experiments on actual lawn grass areas. Now, the suggested algorithm and the fusion of sensors are evaluated against wide range of lawn and grounds.

Digital Twin Driven Requirement Conversion in Smart Customized Design

Requirement conversion (RC) is an important activity in product customized design. Nowadays, the RC process is mostly driven by designers’ knowledge, the RC model is passive to dynamic customer requirements, and the RC behavior is a black box to customers. Digital twin (DT) is characterized as a self-reinforcing mechanism driven by mirroring between the physical and virtual spaces, which is powerful in addressing these challenges for RC. This paper proposes a digital twin driven requirement conversion (DTRC) architecture and a tri-model-based approach integrated by digital model, behavior model, and evolution model for DTRC development. Firstly, the digital model based on the artificial neural network can simulate the virtual twin data to compensate for the absence of real-world data. Then, driven by the virtual-reality integration data, the behavior model mirrors and visualizes the RC behavior in real world based on the decision tree. Finally, a genetic algorithm based evolution model optimizes the RC rules via physical data throughout the whole product life cycle. A case study of DTRC for elevator customized design is further conducted to validate feasibility and effectiveness of the proposed approach. Experimental results show that DTRC outperforms other RC approaches in terms of conversion accuracy. Meanwhile, DTRC can visualize and optimize the conversion path through the tree topology, which is beneficial to the customer participation and proactive to the dynamic environment.

Salvaging COVID‐19 interrupted Alzheimer clinical trials using virtual patient simulations

BackgroundThe COVID‐19 pandemic has halted many Alzheimer’s Disease clinical trials, with some forced to re‐start at different points along the trial time‐line with substantial protocol amendments and patients dropping out with substantial partial assessments. To align the functional outcomes of these protocol amendments with the original trial design at the individual patient level, we propose to use Physiology‐Based Pharmacokinetic (PBPK) and Quantitative Systems Pharmacology (QSP) Modeling to ‘correct’ the cognitive trajectory of mirror virtual patient population back to the original trail design.MethodThe Virtual Twin approach creates a PBPK computer‐simulated model of each patient with a virtual twin QSP model of trial subjects, with the same co‐medications, common genotype variants affecting metabolism and cognitive outcome; β‐amyloid and tau biomarkers. The QSP platform is a previously ADAS‐Cog calibrated model of key neuronal circuits involved in cognition, allowing to model the effects of CNS active co‐medications based on their pharmacology and genotypes based on imaging studies. In this Virtual Twin approach, the platform will be extensively validated against the actual clinical data from the completer set and the fragmented outcomes of the restarters with their individual protocol amendments, before simulating the cognitive trajectory with the original trial design for those whose trial was interrupted.ResultDifferent interruption scenarios in a 24‐month AD study of a bi‐weekly amyloid antibody infusion are simulated with increased anxiolytics and anti‐depressants use after restart. The impact of changes is dependent upon the genotype combination with the 5‐HTTLPR rs 23351 > APOE > COMTVal158Met and different for placebo and active treatment. Anti‐depressants and benzo also differentially substantially affect the cognitive trajectory. We illustrate how to reconstruct the cognitive trajectory of 600 subjects (out of 1200) affected by the interruption.ConclusionCombining QSP modeling of the biology with PBPK modeling and extensive validation with the fragmented clinical data available, in principle allows to reconstruct the original cognitive trajectory in patients affected by the COVID‐19 interruption. In this way, the original trial design and statistical analysis plan can be applied to achieve a fair evaluation of the clinical effect of the investigative new drug.

Vision guided robotic inspection for parts in manufacturing and remanufacturing industry

Environmental and commercial drivers are leading to a circular economy, where systems and components are routinely recycled or remanufactured. Unlike traditional manufacturing, where components typically have a high degree of tolerance, components in the remanufacturing process may have seen decades of wear, resulting in a wider variation of geometries. This makes it difficult to translate existing automation techniques to perform Non-Destructive Testing (NDT) for such components autonomously. The challenge of performing automated inspections, with off-line tool-paths developed from Computer Aided Design (CAD) models, typically arises from the fact that those paths do not have the required level of accuracy. Beside the fact that CAD models are less available for old parts, these parts often differ from their respective virtual models. This paper considers flexible automation by combining part geometry reconstruction with ultrasonic tool-path generation, to perform Ultrasonic NDT. This paper presents an approach to perform custom vision-guided ultrasonic inspection of components, which is achieved through integrating an automated vision system and a purposely developed graphic user interface with a robotic work-cell. The vision system, based on structure from motion, allows creating 3D models of the parts. Also, this work compares four different tool-paths for optimum image capture. The resulting optimum 3D models are used in a virtual twin environment of the robotic inspection cell, to enable the user to select any points of interest for ultrasonic inspection. This removes the need of offline robot path-planning and part orientation for assessing specific locations on a part, which is typically a very time-consuming phase.

Deep learning for brake squeal: Brake noise detection, characterization and prediction

Despite significant advances in modeling of friction-induced vibrations and brake squeal, the majority of industrial research and design is still conducted experimentally, since many aspects of squeal and its mechanisms involved remain unknown. In practice, measurement data is available in large amounts. We report here for the first time on novel strategies for handling data-intensive vibration testings to gain better insights into friction brake system vibrations and noise generation mechanisms. Machine learning-based methods to detect and characterize vibrations, to understand sensitivities and to predict brake squeal are applied with the aim to illustrate how interdisciplinary approaches can leverage the potential of data science techniques for classical mechanical engineering challenges.In the first part, a deep learning brake squeal detector is developed to identify several classes of typical friction noise recordings. The detection method is rooted in recent computer vision techniques for object detection based on convolutional neural networks (CNN). It allows to overcome limitations of classical approaches that solely rely on instantaneous spectral properties of the recorded noise. Results indicate superior detection and characterization quality when compared to a state-of-the-art brake squeal detector. In the second part, a recurrent neural network (RNN) is employed to learn the parametric patterns that determine the dynamic stability of an operating brake system. Given a set of multivariate loading conditions, the RNN learns to predict the noise generation of the structure. The validated RNN represents a virtual twin model for the squeal behavior of a specific brake system. It is found that this model can predict the occurrence and the onset of brake squeal with high accuracy and that it can identify the complicated patterns and temporal dependencies in the loading conditions that drive the dynamical structure into regimes of instability. Large data sets from commercial brake system testing are used to train and validate the models. This work is a contribution to the MSSP Special Issue in Honor of Professor Lothar Gaul.

Transformer hot spot temperature prediction based on basic operator information

A power transformer is an important component in power trains and electrical distribution networks. Predicting its life time is desirable, especially, if a zero downtime policy is applied. However, end customers often have to deal with a lack of information and cannot always use the established methods for life time prediction. Therefore, the present paper provides an alternative way to calculate the hot spot temperature and thus, the life time of power transformers based on limited information, i.e. transformer rating information and rms current and voltage measurements (including phase angles). The transformer hot spot temperature is derived from the transformer losses and a virtual twin. Therefore, the paper provides methods i) to evaluate the separate transformer losses, i.e. core, winding and stray losses, ii) to create a simple virtual transformer twin and iii) to calculate the temperature distribution in the transformer windings and thus, the hot spot temperature. The methods are applied to one phase of a 154 kV, 15MVA power transformer. It is shown that the calculated losses and hot spot temperature matches with winding measurements available in literature.

Salvaging CNS Clinical Trials Halted Due to COVID-19.

The COVID-19 pandemic has halted many ongoing CNS clinical trials, especially in Alzheimer's disease. These long-duration trials involve many stakeholders, especially the patients and their family members, who have demonstrated their commitment to developing new therapeutic interventions for this devastating disease. We certainly do not want to lose all the knowledge we have gained from these ongoing trials because of the pandemic. While some of these trials will need to restart, others can re-start at different points along the trial protocol with substantial protocol amendments. However, there is an urgent need to combine useful information from the completers with those subjects undergoing complex protocols deviations and amendments after re-start. We propose the concept of mechanistic modeling-based virtual twin patients as a possible solution to harmonize the readouts from these complex and fragmented clinical datasets in a biologically relevant way.

A modeling informed quantitative approach to salvage clinical trials interrupted due to COVID-19.

Many ongoing Alzheimer's disease central nervous system clinical trials are being disrupted and halted due to the COVID‐19 pandemic. They are often of a long duration’ are very complex; and involve many stakeholders, not only the scientists and regulators but also the patients and their family members. It is mandatory for us as a community to explore all possibilities to avoid losing all the knowledge we have gained from these ongoing trials. Some of these trials will need to completely restart, but a substantial number can restart after a hiatus with the proper protocol amendments. To salvage the information gathered so far, we need out‐of‐the‐box thinking for addressing these missingness problems and to combine information from the completers with those subjects undergoing complex protocols deviations and amendments after restart in a rational, scientific way. Physiology‐based pharmacokinetic (PBPK) modeling has been a cornerstone of model‐informed drug development with regard to drug exposure at the site of action, taking into account individual patient characteristics. Quantitative systems pharmacology (QSP), based on biology‐informed and mechanistic modeling of the interaction between a drug and neuronal circuits, is an emerging technology to simulate the pharmacodynamic effects of a drug in combination with patient‐specific comedications, genotypes, and disease states on functional clinical scales. We propose to combine these two approaches into the concept of computer modeling‐based virtual twin patients as a possible solution to harmonize the readouts from these complex clinical datasets in a biologically and therapeutically relevant way.

Моделирование работы печи гомогенизации для решения задач управления

The purpose of the work is to create a mathematical model of furnace thermal operation for control problems, on the basis of which a "virtual twin" of the homogenizing furnace can be implemented. Recently, the heating quality indicators of aluminum semi-finished products and products before further processing are becoming particularly relevant in the processes of aging or hardening also due to the introduction of the international quality standard ISO at metallurgical enterprises. At the same time, the existing control systems of heating furnace parameters mainly regulate the local heater by measuring the air temperature near it. The task of maintaining the desired heating volume of the entire furnace and the charge is quite difficult under this type of control and requires constant adjustment of the proportional-integral-differentiating regulators by an experienced employee. The energy intensity of aluminum processing is also high but not all enterprises are able to replace outdated furnaces. Then the resource-saving opportunities center around the improvement of control systems. In particular, the replacement of traditional response regulation with proactive regulation is a promising direction. Therefore, the development of mathematical models and software enabling to calculate and predict the temperatures of different areas of a furnace and heated materials depending on the change in heating conditions is a relevant task. Modeling and studies are carried out using ordinary differential equations and explicit difference schemes. The paper proposes a dynamic model of heat transfer in a homogenizing furnace and performs test calculations demonstrating the applicability of this model to the selection of heating modes and algorithms. The model can also be used as a basis for the development of a furnace digital twin. The presented models and calculation methods are applicable as developments of modern control systems and heating optimization of not only metal products but any materials.

Virtual Engineering of a Fusion Reactor: Application to Divertor Design, Manufacture, and Testing

Virtual engineering is being widely adopted in diverse technical fields and has the potential to accelerate the realization of tokamak fusion energy. In this paper, we demonstrate, for the first time, the use of virtual engineering techniques to enhance the development cycle of a water-cooled divertor plasma facing component. The divertor concept under development is the thermal break, which aims to increase heat flux limits and DEMO-relevance by alleviating the high thermal mismatch stress between a tungsten monoblock and a copper alloy cooling pipe. Response surface-based design search and optimization is used to explore the potential of the concept and to inform the design of a high-heat-flux (HHF) mock-up. In developing a brazing manufacturing procedure, computed tomography scanning and subsequent image-based finite-element modeling revealed a defective joining process which would result in unacceptable divertor performance. The same techniques were used to develop an improved brazing procedure that was used to manufacture the mock-ups. A production defect was discovered by ultrasonic qualification testing, but a “virtual twin” model confirmed the acceptable performance of the mock-up under HHF. Indeed, all mock-ups were then subjected to at least 100 cycles of 20 MW/m 2 HHF testing, with no signs of damage. The development of these virtual engineering and modeling techniques will facilitate criteria for design and component rejection, and in future enable predictive maintenance of components in service. Practical trials and experiments will remain essential to complement and validate simulations.